Despite its potential benefits of reduced traffic congestion and discounted trips, incorporating ride-pooling in a city comes with a set of challenges that require thorough analysis, optimisation, and planning. Even though, services like \textit{Uber} have existed in Amsterdam for over a decade, city wide ride-pooling has yet to be implemented. This paper uses an algorithm for exact matching of attractive shared rides (ExMAS) and Albatross travel demand data to map and analyse the spatial disparities of key performance indicators of a ride-pooling service in Amsterdam and discover the potential of certain areas in the city. The experiments utilised a set of increasing discounted fares for ride-pooling with increasing travel demand levels. A ride-pooling service with higher discounted fares generally increased the attractiveness of the system and reduced the total vehicle hours, when compared to its non-shared counterpart. It was found that the largest vehicle hour reduction were in areas on the periphery of Amsterdam (namely the West, North, and East areas) where rides of higher degree and longer trips lengths were more likely. However, the user attractiveness of the system tended to be higher in central areas of the city where trip density was higher, trip length shorter, and ride degrees lower. The study also determined that variance of the vehicle hours and user attractiveness decreased and stabilised with increasing demand level. This paper could be a starting point in optimising the possible roll out schemes for a ride-pooling service in Amsterdam.

Ride-pooling is a key concept for the future of human mobility and vital in the roll-out of Mobility-as-a-Service (MaaS). Pooling allows individuals to travel at a reduced fare (when comparing to a private alternative) due to the reduction in operating costs as the usage of the vehicles is increased; essentially, the main intentions of pooling are to alleviate traffic congestion, reduce required operator fleet size, and to reduce vehicle hours travelled. Ride-pooling does have its drawbacks, travellers increase their in-vehicle travel times due to the detours induced by sharing. A way of minimising the time lost due to detouring is to incorporate pick-up and drop-off (PUDO) points that travellers would walk to and from when opting for ride-pooling. However, the knowledge on the extent of these benefits with respect to the service configurations used is still limited, therefore the objective of this thesis is to examine how pricing configurations and service settings affect the operator performance and level of service of ride-pooling with PUDO. We extend the utility formulation of an existing algorithm that matches trip requests to attractive shared rides where a route search algorithm assesses a PUDO configuration of a ride by computing and comparing the utility of the vehicle and the utilities of the travellers within the vehicle. The algorithm is applied to the context of Amsterdam and aims to further optimise selected pooled rides where experimentation consists of varying the door-to-door pooling discount, the PUDO discount, the service setting, and the demand level. The service setting sets weights on the utilities where we can favour the vehicle during the route search or treating the travellers and vehicle equally. Results show that increasing PUDO discount increases general attractiveness of the service allowing for more travellers to opt for ride-pooling with PUDO, however the largest differences in system-wide performance occurred when PUDO discount was significantly larger than door-to-door discount. Total vehicle hours could be reduced up to 2.2%, improve passenger utility by 2.8% but can suffer a loss of revenue up to 11.4%. The service setting was also able to control these performance indicators as favouring the vehicle provided the largest reductions in vehicle hours and lowest loss in revenue while treating the traveller and vehicle equally was able to provide the largest improvement in traveller utility. The former service setting induces longer walking times on travellers which is the cause of the greater reductions in vehicle travel time while the latter service setting is the opposite and the reason to why travellers find it more attractive. In essence, ride-pooling with PUDO is able to further reduce vehicle hours and improve traveller utility; pricing configurations and service settings can be helpful with scenarios where supply exceeds demand and vice versa. The use of the service setting showed that ride-pooling with PUDO can be made much more attractive to travellers by setting fair PUDO points to walk to by sacrificing vehicle travel time savings. Such a traveller orientated service setting could be useful when supply exceeds demand. A service provider such as Uber could utilise the insights obtained from this thesis when rolling out such a service in Amsterdam and plan for certain scenarios.